ports/openmpi
ports/openmpi
1
1
Форкнуть 0
Код Релизы Активность
7 838 коммитов 3 Ветки 0 Теги
Граф коммитов

1122 Коммитов

Автор SHA1 Сообщение Дата
George Bosilca
27000ef7d6 More compact and readable code. Otherwise, no big difference with the 
previous version.

This commit was SVN r10389.
 2006-06-16 03:07:42 +00:00
George Bosilca
3f96f39e46 If the goal of this code was to copy the iovec and skip the first offset 
bytes then it was not correct.

This commit was SVN r10388.
 2006-06-16 03:06:30 +00:00
George Bosilca
93afe59226 It is not required to initialize the csum. 
This commit was SVN r10387.
 2006-06-16 03:05:20 +00:00
George Bosilca
1f96768b76 For zero length persistent request do not reposition the convertor as 
it is not initialized.

This commit was SVN r10386.
 2006-06-16 03:04:41 +00:00
Brian Barrett
05046e8ad2 if MX isn't running on some hosts, but is on others, we were blocking in the modex receive 
waiting for the non-running procs to publish their contact information.  Publish their
(lack of) contact information.

This commit was SVN r10355.
 2006-06-14 19:07:38 +00:00
George Bosilca
aca71521db Complete the move of the mpool registration from opal_list_item_t to the 
ompi_free_list_item_t.

This commit was SVN r10354.
 2006-06-14 17:43:50 +00:00
Galen Shipman
5d71c149c2 Another fix for PML request completion when local network completion can occur 
out of order.. 

Reviewed by Brian.. needs to hit 1.1 

This commit was SVN r10353.
 2006-06-14 16:55:35 +00:00
Brian Barrett
d367dc5d56 * Fix for bug #115 -- we need to decrement the use count on a pinned buffer 
so that memory is actually deregistered.  Reviewed by Galen.

This commit was SVN r10349.
 2006-06-14 13:38:24 +00:00
George Bosilca
3727fa2ae6 Nothing relevant. I add some more output in the case we have a checksum error. 
Just to be able to know more information about the failure.

This commit was SVN r10337.
 2006-06-13 19:36:38 +00:00
Galen Shipman
0eddad6849 Handle out of order completion/receives when marking completion... 
this is a fix for #107... needs to go to the 1.1 branch.. 

This commit was SVN r10331.
 2006-06-13 16:57:41 +00:00
Andrew Friedley
c68c6ac122 A number of fixes and the usual cleanup.. 
- Added some basic flow control to limit number of posted sends.
- Merged endpoint send/recv lock into single endpoint lock.
- Set the LMR triplet length in the send path, not at allocation time.
  This has to be done because upper layers might send less than the
  amount allocated.
- Alter the tie-breaker if statement protecting the second call
  to dat_ep_connect().  The logic was reversed compared to the tie-
  breaker for the first dat_ep_connect(), making it possible for
  3 or more processes to form a deadlock loop.
- Some asserts were added for debugging purposes.. leaving them
  in place for now.

This commit was SVN r10317.
 2006-06-12 22:42:01 +00:00
Galen Shipman
218a438509 finished the ompi_free_list_t class nightmare.. 
This commit was SVN r10314.
 2006-06-12 22:09:03 +00:00
Galen Shipman
18dda70fd0 make ompi_free_list_item_t a class.. 
This will go to the 1.1 branch but will probably require a few changes as
ompi_free_list_t is different in the branch.. 

This commit was SVN r10306.
 2006-06-12 16:44:00 +00:00
Brian Barrett
d3257f22d8 * back out Galen's r10300 because it breaks the build. Real fix coming RSN. 
This commit was SVN r10303.

The following SVN revision numbers were found above:
  r10300 --> open-mpi/ompi@b0f3745791
 2006-06-12 14:38:14 +00:00
Gleb Natapov
48d348b577 Don't complete send request before we've got completion on the first rndv packet. 
Sender can receive and complete PUT request before it gets completion on the first rndv packet. senreq struct may be reused for the next MPI_Send and unexpected completion mess up the things. I sometimes got SEGV and sometimes data corruption.

This commit was SVN r10301.
 2006-06-12 14:00:43 +00:00
Galen Shipman
b0f3745791 declare these as ompi_free_list_item_t's 
This needs to go to 1.1

This commit was SVN r10300.
 2006-06-12 13:26:15 +00:00
George Bosilca
7d1feffbf7 The real solution. If the sendreq->req_send.req_bytes_packed is zero then there 
is no data to be trasfered. And this is the condition which lead to a non
initialized convertor.

This commit was SVN r10299.
 2006-06-12 06:18:18 +00:00
George Bosilca
c959c2f214 Don't reset the convertor's position if it wasn't initialized before. This can 
only happens for zero byte persistent requests.

This commit was SVN r10298.
 2006-06-12 06:14:35 +00:00
Galen Shipman
9d73217637 These list items are free list items, and should inherit properly.. 
This commit was SVN r10295.
 2006-06-11 20:19:12 +00:00
Brian Barrett
d5acb4e3cc * silence dumb (and mostly useless) warning during cleanup 
This commit was SVN r10280.
 2006-06-09 21:09:53 +00:00
Brian Barrett
cc99a63169 * fix issue with PANFS not building properly - we didn't add PANFS_LIB to the 
list of libraries

This commit was SVN r10279.
 2006-06-09 20:41:12 +00:00
Jeff Squyres
a4030ad2d9 Improve the tremendously unhelpful MCA help message for the 
btl_openib_ib_mtu and btl_mvapi_ib_mtu MCA params by showing the valid
values what what they represent (got a question about this from Cisco
testing engineers).

This commit was SVN r10277.
 2006-06-09 18:02:45 +00:00
Andrew Friedley
9a92394bfd Mostly cleanups - preprocessor fixes and removal of OPAL_OUTPUTs. 
Also updated to match recent mpool_free changes.

This commit was SVN r10273.
 2006-06-09 00:18:29 +00:00
Andrew Friedley
75176370ae blah. somehow missed adding .ompi_ignore/.ompi_unignore. 
This commit was SVN r10272.
 2006-06-09 00:15:36 +00:00
Andrew Friedley
cca1616368 Finally committing the UD BTL. 
UD is the Unreliable Datagram transport for Infiniband, specifically OpenIB.  This BTL is derived from the existing openib BTL, which is RC (Reliable Connection) based.

Still a work in progress, as there is a lot of work left to do.  Specifically, performance, scalability, and flow control need to be addressed.

Currently I'm playing around with different methods for handling receive buffers, as well as profiling to figure out where the time is going.

This commit was SVN r10271.
 2006-06-09 00:13:45 +00:00
Galen Shipman
08823e56fa check address before looking for the item in the tree corresponding to the 
address.. 
All have been reviewed by brian.. putting in a changeset request.. 

This commit was SVN r10256.
 2006-06-08 16:27:59 +00:00
Galen Shipman
636ef0cf6c don't put back null items on the list.. 
This commit was SVN r10253.
 2006-06-08 14:46:41 +00:00
Galen Shipman
429056078a fix numerous late night errors.. 
1) don't need tree if memory is just malloc'd 
2) fix memory and free list leak.. 
3) deregister first and then free... doh.. 

This commit was SVN r10251.
 2006-06-08 14:23:20 +00:00
Galen Shipman
5a2ceda93f a couple of stupid late night mistakes... 
This commit was SVN r10250.
 2006-06-08 13:39:41 +00:00
Galen Shipman
0bb8a6fca8 roll back to not use memalign 
This commit was SVN r10249.
 2006-06-08 04:34:04 +00:00
Galen Shipman
b42b0bd1af potential fix for ticket #81 
Added a tree to track memory allocation from MPI_Alloc_mem, this allows us to
free the registrations in a sane fashion.. also should be faster.. 

This commit was SVN r10248.
 2006-06-08 04:29:27 +00:00
Sven Stork
c31e6f9767 use memalign instead of malloc + manually alignment in the mvapi mpool 
revert commit 10243

This commit was SVN r10247.
 2006-06-07 23:21:23 +00:00
Andrew Friedley
5ace292cc1 Should fix ticket #81 - which is specific to MVAPI, I've included the same fix for gm/openib as well. 
uDAPL has the same problem, will fix in separate commit so it doesn't go to branch.

This commit was SVN r10243.
 2006-06-07 15:52:48 +00:00
Galen Shipman
84479d0b5a potential fix for iprobe test,, tested with openib.. will have andy try ud.. 
This commit was SVN r10232.
 2006-06-06 22:10:41 +00:00
Galen Shipman
90799f82cd copy paste error.. 
This commit was SVN r10220.
 2006-06-06 02:38:29 +00:00
Galen Shipman
cc54b07aa0 add better error messages for vapi retry exceeded errors. 
This commit was SVN r10219.
 2006-06-06 02:04:56 +00:00
Galen Shipman
9e6e7575b9 doh... add the file.. 
This commit was SVN r10210.
 2006-06-05 21:24:42 +00:00
Galen Shipman
f05dee0435 add help file to explain why things went south.. 
This commit was SVN r10209.
 2006-06-05 21:23:45 +00:00
Galen Shipman
74c97fb784 cleanup error reporting.. use ompi_proc_t->proc_name if available this gives 
us source/dest hostnames for communication errors.. 

This goes to 1.1 branch (reviewed by Brian).. 

This commit was SVN r10200.
 2006-06-05 20:02:41 +00:00
Brian Barrett
c70fff6ed0 * Fix for bug #44 for the trunk -- remove a bunch of warnings from the DR 
PML when compiling on Solaris.  Patch won't apply cleanly to the v1.1
  branch, so a diff for that is coming up soon.

This commit was SVN r10173.
 2006-06-01 18:58:38 +00:00
Galen Shipman
83ff3201b5 don't use rank or nprocs in error messages when we don't have them.. 
This should hit 1.1 and 1.0 branches.. 
Reviewed by Brian

This commit was SVN r10164.
 2006-06-01 14:24:11 +00:00
Galen Shipman
0344ae4ac5 Fix to allow eager limit and max send size to be any size (within resource limitations). Instead of storing the ompi_free_list_t * in the fragment, we use the frag type enum, this tells us where the frag came from and where it should return.. This could also be done in mvapi but is not a high priority moving forward.. 
Review by Brian, needs to hit the trunk + 1.1 release.. 

This commit was SVN r10157.
 2006-06-01 02:32:18 +00:00
Brian Barrett
5163f2b296 Fix for bug #36. The MX, MVAPI, and OpenIB components don't have 
support for progress threads, so we shouldn't build them or try to use
them when support for progress threads has been requested.  The TCP, GM,
SELF, and SM BTLs should have progress thread support, so they aren't
disabled.  The Portals BTL isn't compiled on platforms with threads,
so it doens't need to be updated.

This commit was SVN r10156.
 2006-06-01 01:30:16 +00:00
Galen Shipman
c79efc9efb track which list a fragment came from, allows returning based on list, not 
on size. 

This commit was SVN r10142.
 2006-05-31 14:24:32 +00:00
Brian Barrett
4904e34a52 set datarootdir, necessary for Autoconf-2.60 which will define some variables 
based upon this value (e.g., datadir, docdir).

Submitted by: Ralf Wildenhues
Reviewed by: Brian Barrett

This commit was SVN r10133.
 2006-05-31 03:43:55 +00:00
Brian Barrett
6026fc98f6 * Fix M4 quoting so that AC 2.60 won't complain 
Submitted by: Ralf Wildenhues
Reviewed by: Brian Barrett

This commit was SVN r10129.
 2006-05-31 03:39:18 +00:00
Brian Barrett
c723d196c5 Rather than using fragment size to determine fragment type, use an enum. 
Do this rather than the my_list pointer because we need to do some
things that are somewhat special because we pre-pin eager fragments but
not send fragments.  Also makes a couple ideas I have slightly easier to
play around with.

This commit was SVN r10127.
 2006-05-31 03:34:32 +00:00
Galen Shipman
2667c52a5d Track fragments by list, not by size.. 
-- reviewed by Brian, needs to hit all the branches.. 

This commit was SVN r10078.
 2006-05-25 18:07:26 +00:00
Galen Shipman
38a0561d9b Allow maximum send size to be less than the eager limit. 
Instead of figuring out which free list the fragment belongs to based on size
we simply store a pointer to the list which it belongs in the fragment.

This was reviewed by Brian and should hit all the branches.

This commit was SVN r10072.
 2006-05-25 16:57:14 +00:00
Andrew Friedley
fa9ec2afdf Add my sandia username for convenience 
This commit was SVN r10071.
 2006-05-25 15:49:11 +00:00